Gait recognition based on MEMS accelerometer

Recently, with the rapid development of MEMS technology, the micro-sensor´s application in gait research has become more and more widespread, relying on its small size, low cost, light weight, and high precision characteristics. This paper presents a non-specific human gait type recognition system based on a single MEMS accelerometer, and will guide the further study in human identification, motion analysis, medical care, diet plans, etc. After a series of computing and processing the 3D raw acceleration data, by using the wavelet-threshold algorithm, the signal and noise are separated and gait cycles are easily divided. By adopting the pattern recognition theory and the combination method of time domain and frequency domain, the system realizes the gait type recognition. In the experiment the EER can reach 6.29% with 400 sets of acceleration data for test.